https://www.interviewquery.com/p/data-science-job-market-disappearing
so we looked at the data
The Data Science Job Market is Disappearing
-
The skills needed in data science aren't going anywhere. It's just the supply of those skills have increased so their wage premium has dropped. Earlier if you wanted someone to occasionally run logistic regression you'd advertise for a "data scientist" job and people who had superficial knowledge of all these machine learning methods would apply. Now that many know how to run and interpret a simple regression you'll simply hire an analyst and not ask for a "data scientist".
-
The skills needed in data science aren't going anywhere. It's just the supply of those skills have increased so their wage premium has dropped. Earlier if you wanted someone to occasionally run logistic regression you'd advertise for a "data scientist" job and people who had superficial knowledge of all these machine learning methods would apply. Now that many know how to run and interpret a simple regression you'll simply hire an analyst and not ask for a "data scientist".
when you say "superficial knowledge [of these methods]", even world-renowned experts like Andrew Ng don't have much intuition for why certain models work. He says that himself a few times in the DeepLearning.ai course sequence.
-
The skills needed in data science aren't going anywhere. It's just the supply of those skills have increased so their wage premium has dropped. Earlier if you wanted someone to occasionally run logistic regression you'd advertise for a "data scientist" job and people who had superficial knowledge of all these machine learning methods would apply. Now that many know how to run and interpret a simple regression you'll simply hire an analyst and not ask for a "data scientist".
when you say "superficial knowledge [of these methods]", even world-renowned experts like Andrew Ng don't have much intuition for why certain models work. He says that himself a few times in the DeepLearning.ai course sequence.
I guess. But it would be fair to say that we can draw some sort of a distinction between the knowledge that Andrew ng has and the knowledge someone who completed a zero-prerequite 1 month course era course on "deep learning" has (but who hasn't done a single course in probability).
Both might have a machine learning project on their CV using a similar method but one simply knows a python command while the other has a much deeper knowledge.
Not meant to sound c0cky however. Economists are closer to the 1 month course dude than they are to Andrew ng
-
The skills needed in data science aren't going anywhere. It's just the supply of those skills have increased so their wage premium has dropped. Earlier if you wanted someone to occasionally run logistic regression you'd advertise for a "data scientist" job and people who had superficial knowledge of all these machine learning methods would apply. Now that many know how to run and interpret a simple regression you'll simply hire an analyst and not ask for a "data scientist".
when you say "superficial knowledge [of these methods]", even world-renowned experts like Andrew Ng don't have much intuition for why certain models work. He says that himself a few times in the DeepLearning.ai course sequence.
I guess. But it would be fair to say that we can draw some sort of a distinction between the knowledge that Andrew ng has and the knowledge someone who completed a zero-prerequite 1 month course era course on "deep learning" has (but who hasn't done a single course in probability).
Both might have a machine learning project on their CV using a similar method but one simply knows a python command while the other has a much deeper knowledge.
Not meant to sound c0cky however. Economists are closer to the 1 month course dude than they are to Andrew ngProbably not much difference between top econometricians and Andrew Ng really. What you say is true of the run of the mill reg moneky but that's a very strange comparison to make.
-
The skills needed in data science aren't going anywhere. It's just the supply of those skills have increased so their wage premium has dropped. Earlier if you wanted someone to occasionally run logistic regression you'd advertise for a "data scientist" job and people who had superficial knowledge of all these machine learning methods would apply. Now that many know how to run and interpret a simple regression you'll simply hire an analyst and not ask for a "data scientist".
when you say "superficial knowledge [of these methods]", even world-renowned experts like Andrew Ng don't have much intuition for why certain models work. He says that himself a few times in the DeepLearning.ai course sequence.
I guess. But it would be fair to say that we can draw some sort of a distinction between the knowledge that Andrew ng has and the knowledge someone who completed a zero-prerequite 1 month course era course on "deep learning" has (but who hasn't done a single course in probability).
Both might have a machine learning project on their CV using a similar method but one simply knows a python command while the other has a much deeper knowledge.
Not meant to sound c0cky however. Economists are closer to the 1 month course dude than they are to Andrew ngProbably not much difference between top econometricians and Andrew Ng really. What you say is true of the run of the mill reg moneky but that's a very strange comparison to make.
I meant the typical economist that goes into industry
-
I took a probability course years ago. Borel-Cantelli I & II, Kolmogorov 0-1, Cauchy-Schwartz, Chebyshev, MCT, DCT, Fatou, Fubini, etc. I don't remember much tbh and couldn't derive the results now from general knowledge.
Anyway, it's a bit removed from understanding how to implement Deep Learning methods to solve practical problems in CV, NLP, and RecSys. Python fluency is far more important at the practitioner level- which is where Data Science is.
-
I meant the typical economist that goes into industry
so you consider Andrew Ng the typical computer scientist who goes to industry? You are comparing apples and oranges, why not pick some of the chief economists who have gone to tech as a comparison, that would be much fairer.
-
I took a probability course years ago. Borel-Cantelli I & II, Kolmogorov 0-1, Cauchy-Schwartz, Chebyshev, MCT, DCT, Fatou, Fubini, etc. I don't remember much tbh and couldn't derive the results now from general knowledge.
Anyway, it's a bit removed from understanding how to implement Deep Learning methods to solve practical problems in CV, NLP, and RecSys. Python fluency is far more important at the practitioner level- which is where Data Science is.Huh, I remember those times where I took grad level probability and functional analysis, but I do not look fondly on them at all... It's sad how I have mostly forgotten about them, but at the same time I'm glad to be employed
-
when you say "superficial knowledge [of these methods]", even world-renowned experts like Andrew Ng don't have much intuition for why certain models work. He says that himself a few times in the DeepLearning.ai course sequence.
The funny thing about the industry is that they really don't care about intuition and/or details about why a model works. Indeed, industry interview cares about lower level skill like Leetcoding rather than your ability to provide intuition / proofs for a model. Heck, if Andrew Ang himself were to apply for data scientist jobs, and he removes his name from his CV, he'd probably flunk out of the interview because he couldn't answer some BS Leetcode question.